Home / Product / Atlas and Studio

C17 - Hermes "Send to Studio" (Hermes creates a Studio draft) - build brief (v1, 2026-06-20)

Updated Jun 20, 2026 · Affirmology_SendToStudio_C17_BuildBrief_v1.md

C17 - Hermes "Send to Studio" (Hermes creates a Studio draft) - build brief (v1, 2026-06-20)

GREENLIT by Jeff 2026-06-20 (Hermes flagged it can't create a Studio entry / trigger a render from chat). This closes the gap that makes Hermes a writing assistant instead of a production assistant. Same family as C16 (cross-chat push) and C15 (field-briefing audio): Hermes authoring something, then ACTING on it across the toggle. Build it in the same pass as C16 - identical shape (an API endpoint + a Hermes tool + a chat button), so the plumbing is shared.

What it does

One "Send to Studio" button (and the spoken equivalent, "send this to the Studio") in the Hermes chat takes the script Hermes just wrote and lands it in the Studio as a DRAFT entry, pre-populated: person, structure, music bed, production notes, and the draft script. The user switches to the Studio and it's sitting there ready to render or regenerate. No copy-paste, no rebuilding context.

Architecture note (keep the chart-driven rule intact)

EVERYTHING IS CHART-DRIVEN and the Studio is where the canonical render happens. So the draft carries the PARAMETERS the Studio needs (person, structure/template, length, music bed, and Hermes's production directives) PLUS Hermes's draft script as an editable, regenerable field. The user can render Hermes's draft as-is or hit regenerate to have the Studio re-fill the script from the chart + structure. Hermes's script is itself chart-derived (it was written for that person's chart), so this complies; we are not introducing chartless content, and the Studio stays the place the final render is produced.

Data model

Reuse the Studio's EXISTING entry / render-queue model. Add (or set) a status = "draft" (un-rendered) state and a source = "hermes" marker. Fields to populate: person, structure, length, music_bed, production_notes, draft_script.

API (api/main.py), bearer-auth

• POST /api/studio/draft body {as_person, person, structure, length?, music_bed?, production_notes?, draft_script} -> create a draft Studio entry via the existing entry-creation path, status "draft", source "hermes" -> {ok:true, entry_id}.
• The Studio entry list / studio.html should show draft entries (with a "from Hermes" tag) alongside rendered ones, each with a Render action.

Hermes tool (api/hermes_tools.py)

Add send_to_studio to HERMES_TOOLS + dispatch: {person, structure, length?, music_bed?, production_notes?, draft_script} -> same insert as the endpoint. Tool description MUST state: only call when the user explicitly asks to send/push the script to the Studio. Hermes never does it on its own initiative (user-initiated action, not autonomous).

UI (web/app.js + studio.html)

• A "Send to Studio" control on a Hermes script/reading -> POST /api/studio/draft, then a confirmation with a hint to switch to the Studio toggle.
• studio.html entry list shows draft entries (status "draft", "from Hermes" tag) with a Render button. Clean copy, no em dashes.

Permission tiers (Atlas spec)

• jeff: full (create drafts freely).
• sol, colin: may create drafts for themselves; reduced per the Atlas spec; never overwrite/delete existing entries (drafts are additive).

MVP vs later

• MVP: create the draft entry from a Hermes script (person + structure + music bed + notes + draft script), visible in the Studio list, render-ready. User still clicks Render in the Studio.
• LATER: Hermes triggers the render itself ("render it now") behind a confirm; template auto-selection from the reading; draft -> render status callbacks surfaced back in the chat.

Guardrails

• ADDITIVE ONLY: never overwrite or delete existing Studio entries. A draft is a new entry.
• USER-INITIATED ONLY: Hermes creates a draft only on an explicit user request; it does not auto-create or auto-render.
• DEMO UNTOUCHED: this is a Studio-side draft path; the locked demo render settings are not affected.
• CHART-DRIVEN preserved (see architecture note). No em dashes. Tiers enforced server-side.

Acceptance test

1. In Jeff's Hermes session, after a script: "Send this to the Studio." Hermes calls send_to_studio, returns confirmation with the entry id.
2. Open the Studio: a draft entry tagged "from Hermes" appears, pre-populated with the script, structure, and music bed.
3. Hit Render: it renders through the normal Studio path (QC runs, etc.).
4. Permission: Sol can create her own draft; no one can overwrite/delete an existing entry via this path.